The Republic of Korea plans to launch a lunar orbiter and lander by 2020. There are several ways to enter lunar orbit: direct transfer, phasing loop transfer, weak stability boundary transfer, and spiral transfer trajectory. In this study, trajectory optimization is investigated for a lunar orbiter using a pattern search method that minimizes the required delta-V for direct lunar transfer. This method generates neighborhood points near the initial condition and then determines whether there is a new point that can reduce the value of the objective function. Classical methods require the gradient and acceleration of the objective function, but pattern search does not. Six poll methods and nine search methods are chosen; thus, 54 combinations of poll and search methods are available. The pattern search method can reduce the required delta-V on average by a few meters per second for a time of flight of five days and more than 10 m/s for a time of flight of four or six days, regardless of whether translunar injection is performed at the ascending or descending node.